math_project2¶

I. Problem Statement¶

The purpose of this project is to develop a YOLOV5 license plate detection model that can automatically localize license plates in images with high accuracy and efficiency. License plate detection is important in various applications, including traffic management, parking management, toll collection, law enforcement, and surveillance. Manual detection of license plates is a challenging task due to the large number of vehicles on the road and the variability in license plate design and positioning. Manual detection also requires a significant amount of time and resources, which can be a bottleneck in large-scale applications.

II. Methodology¶

Data collection and preprocessing:

https://www.kaggle.com/datasets/andrewmvd/car-plate-detection

The 433 images with bounding box annotations which is collected on kaggle will be used for training and testing the model. We will preprocessing the dataset by Resize the images, Normalize the pixel values and Extract the license plate region.

Group Members¶

Jincheng Jiang Zhuang Miao

Clone and install the YOLO from github repo

In [1]:
!git clone https://github.com/ultralytics/yolov5
%cd yolov5
!pip install -qr requirements.txt

Cloning into 'yolov5'...
remote: Enumerating objects: 15365, done.
remote: Counting objects: 100% (10/10), done.
remote: Compressing objects: 100% (10/10), done.
remote: Total 15365 (delta 1), reused 2 (delta 0), pack-reused 15355
Receiving objects: 100% (15365/15365), 14.36 MiB | 28.39 MiB/s, done.
Resolving deltas: 100% (10504/10504), done.
/kaggle/working/yolov5
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
dask-cudf 21.12.2 requires cupy-cuda115, which is not installed.
cudf 21.12.2 requires cupy-cuda115, which is not installed.
tensorflow 2.11.0 requires protobuf<3.20,>=3.9.2, but you have protobuf 3.20.3 which is incompatible.
tensorflow-transform 1.12.0 requires pyarrow<7,>=6, but you have pyarrow 5.0.0 which is incompatible.
tensorflow-serving-api 2.11.0 requires protobuf<3.20,>=3.9.2, but you have protobuf 3.20.3 which is incompatible.
librosa 0.10.0.post2 requires soundfile>=0.12.1, but you have soundfile 0.11.0 which is incompatible.
distributed 2021.11.2 requires dask==2021.11.2, but you have dask 2022.2.0 which is incompatible.
dask-cudf 21.12.2 requires dask<=2021.11.2,>=2021.11.1, but you have dask 2022.2.0 which is incompatible.
cloud-tpu-client 0.10 requires google-api-python-client==1.8.0, but you have google-api-python-client 2.82.0 which is incompatible.
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
In [2]:
!sudo apt-get update

Get:1 http://packages.cloud.google.com/apt gcsfuse-focal InRelease [5002 B]
Get:2 https://packages.cloud.google.com/apt cloud-sdk InRelease [6361 B]       
Get:3 https://packages.cloud.google.com/apt google-fast-socket InRelease [5015 B]
Get:4 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64  InRelease [1581 B]
Get:5 http://security.ubuntu.com/ubuntu focal-security InRelease [114 kB]      
Get:6 https://packages.cloud.google.com/apt cloud-sdk/main amd64 Packages [416 kB]
Hit:7 http://archive.ubuntu.com/ubuntu focal InRelease                         
Get:8 http://archive.ubuntu.com/ubuntu focal-updates InRelease [114 kB]        
Get:9 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64  Packages [969 kB]
Get:10 http://security.ubuntu.com/ubuntu focal-security/restricted amd64 Packages [2060 kB]
Get:11 http://archive.ubuntu.com/ubuntu focal-backports InRelease [108 kB]
Get:12 http://security.ubuntu.com/ubuntu focal-security/universe amd64 Packages [1027 kB]
Get:13 http://security.ubuntu.com/ubuntu focal-security/main amd64 Packages [2590 kB]
Get:14 http://archive.ubuntu.com/ubuntu focal-updates/main amd64 Packages [3069 kB]
Get:15 http://archive.ubuntu.com/ubuntu focal-updates/restricted amd64 Packages [2198 kB]
Get:16 http://archive.ubuntu.com/ubuntu focal-updates/universe amd64 Packages [1323 kB]
Get:17 http://archive.ubuntu.com/ubuntu focal-backports/main amd64 Packages [55.2 kB]
Get:18 http://archive.ubuntu.com/ubuntu focal-backports/universe amd64 Packages [28.6 kB]
Fetched 14.1 MB in 2s (6161 kB/s)                       
Reading package lists... Done
In [3]:
!sudo apt-get install python3-venv --yes

Reading package lists... Done
Building dependency tree       
Reading state information... Done
The following additional packages will be installed:
  libpython3.8 libpython3.8-minimal libpython3.8-stdlib python3.8
  python3.8-minimal python3.8-venv
Suggested packages:
  python3.8-doc binfmt-support
The following NEW packages will be installed:
  python3-venv python3.8-venv
The following packages will be upgraded:
  libpython3.8 libpython3.8-minimal libpython3.8-stdlib python3.8
  python3.8-minimal
5 upgraded, 2 newly installed, 0 to remove and 73 not upgraded.
Need to get 6314 kB of archives.
After this operation, 46.1 kB of additional disk space will be used.
Get:1 http://archive.ubuntu.com/ubuntu focal-updates/main amd64 libpython3.8 amd64 3.8.10-0ubuntu1~20.04.7 [1626 kB]
Get:2 http://archive.ubuntu.com/ubuntu focal-updates/main amd64 python3.8 amd64 3.8.10-0ubuntu1~20.04.7 [387 kB]
Get:3 http://archive.ubuntu.com/ubuntu focal-updates/main amd64 libpython3.8-stdlib amd64 3.8.10-0ubuntu1~20.04.7 [1675 kB]
Get:4 http://archive.ubuntu.com/ubuntu focal-updates/main amd64 python3.8-minimal amd64 3.8.10-0ubuntu1~20.04.7 [1903 kB]
Get:5 http://archive.ubuntu.com/ubuntu focal-updates/main amd64 libpython3.8-minimal amd64 3.8.10-0ubuntu1~20.04.7 [717 kB]
Get:6 http://archive.ubuntu.com/ubuntu focal-updates/universe amd64 python3.8-venv amd64 3.8.10-0ubuntu1~20.04.7 [5452 B]
Get:7 http://archive.ubuntu.com/ubuntu focal/universe amd64 python3-venv amd64 3.8.2-0ubuntu2 [1228 B]
Fetched 6314 kB in 1s (7724 kB/s)      
(Reading database ... 111522 files and directories currently installed.)
Preparing to unpack .../0-libpython3.8_3.8.10-0ubuntu1~20.04.7_amd64.deb ...
Unpacking libpython3.8:amd64 (3.8.10-0ubuntu1~20.04.7) over (3.8.10-0ubuntu1~20.04.6) ...
Preparing to unpack .../1-python3.8_3.8.10-0ubuntu1~20.04.7_amd64.deb ...
Unpacking python3.8 (3.8.10-0ubuntu1~20.04.7) over (3.8.10-0ubuntu1~20.04.6) ...
Preparing to unpack .../2-libpython3.8-stdlib_3.8.10-0ubuntu1~20.04.7_amd64.deb ...
Unpacking libpython3.8-stdlib:amd64 (3.8.10-0ubuntu1~20.04.7) over (3.8.10-0ubuntu1~20.04.6) ...
Preparing to unpack .../3-python3.8-minimal_3.8.10-0ubuntu1~20.04.7_amd64.deb ...
Unpacking python3.8-minimal (3.8.10-0ubuntu1~20.04.7) over (3.8.10-0ubuntu1~20.04.6) ...
Preparing to unpack .../4-libpython3.8-minimal_3.8.10-0ubuntu1~20.04.7_amd64.deb ...
Unpacking libpython3.8-minimal:amd64 (3.8.10-0ubuntu1~20.04.7) over (3.8.10-0ubuntu1~20.04.6) ...
Selecting previously unselected package python3.8-venv.
Preparing to unpack .../5-python3.8-venv_3.8.10-0ubuntu1~20.04.7_amd64.deb ...
Unpacking python3.8-venv (3.8.10-0ubuntu1~20.04.7) ...
Selecting previously unselected package python3-venv.
Preparing to unpack .../6-python3-venv_3.8.2-0ubuntu2_amd64.deb ...
Unpacking python3-venv (3.8.2-0ubuntu2) ...
Setting up libpython3.8-minimal:amd64 (3.8.10-0ubuntu1~20.04.7) ...
Setting up python3.8-minimal (3.8.10-0ubuntu1~20.04.7) ...
Setting up libpython3.8-stdlib:amd64 (3.8.10-0ubuntu1~20.04.7) ...
Setting up python3.8 (3.8.10-0ubuntu1~20.04.7) ...
Setting up libpython3.8:amd64 (3.8.10-0ubuntu1~20.04.7) ...
Setting up python3.8-venv (3.8.10-0ubuntu1~20.04.7) ...
Setting up python3-venv (3.8.2-0ubuntu2) ...
Processing triggers for libc-bin (2.31-0ubuntu9.9) ...
Processing triggers for man-db (2.9.1-1) ...
Processing triggers for mime-support (3.64ubuntu1) ...

Build a virtual environment to run yolov5

In [4]:
!python3 -m venv yolov5-env
In [5]:
!source yolov5-env/bin/activate
In [6]:
!pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/cu111/torch_stable.html
!pip install wandb
!pip install pandas
!pip install matplotlib

Looking in links: https://download.pytorch.org/whl/cu111/torch_stable.html
Collecting torch==1.9.0+cu111
  Downloading https://download.pytorch.org/whl/cu111/torch-1.9.0%2Bcu111-cp37-cp37m-linux_x86_64.whl (2041.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.0/2.0 GB 466.5 kB/s eta 0:00:000:01m00:01
Collecting torchvision==0.10.0+cu111
  Downloading https://download.pytorch.org/whl/cu111/torchvision-0.10.0%2Bcu111-cp37-cp37m-linux_x86_64.whl (23.2 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 23.2/23.2 MB 52.9 MB/s eta 0:00:0000:0100:01
Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from torch==1.9.0+cu111) (4.4.0)
Requirement already satisfied: numpy in /opt/conda/lib/python3.7/site-packages (from torchvision==0.10.0+cu111) (1.21.6)
Requirement already satisfied: pillow>=5.3.0 in /opt/conda/lib/python3.7/site-packages (from torchvision==0.10.0+cu111) (9.4.0)
Installing collected packages: torch, torchvision
  Attempting uninstall: torch
    Found existing installation: torch 1.13.0
    Uninstalling torch-1.13.0:
      Successfully uninstalled torch-1.13.0
  Attempting uninstall: torchvision
    Found existing installation: torchvision 0.14.0
    Uninstalling torchvision-0.14.0:
      Successfully uninstalled torchvision-0.14.0
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
pytorch-lightning 1.9.4 requires torch>=1.10.0, but you have torch 1.9.0+cu111 which is incompatible.
kornia 0.6.10 requires torch>=1.9.1, but you have torch 1.9.0+cu111 which is incompatible.
Successfully installed torch-1.9.0+cu111 torchvision-0.10.0+cu111
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: wandb in /opt/conda/lib/python3.7/site-packages (0.14.0)
Requirement already satisfied: protobuf!=4.21.0,<5,>=3.12.0 in /opt/conda/lib/python3.7/site-packages (from wandb) (3.20.3)
Requirement already satisfied: pathtools in /opt/conda/lib/python3.7/site-packages (from wandb) (0.1.2)
Requirement already satisfied: appdirs>=1.4.3 in /opt/conda/lib/python3.7/site-packages (from wandb) (1.4.4)
Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from wandb) (4.4.0)
Requirement already satisfied: PyYAML in /opt/conda/lib/python3.7/site-packages (from wandb) (6.0)
Requirement already satisfied: setuptools in /opt/conda/lib/python3.7/site-packages (from wandb) (67.6.1)
Requirement already satisfied: GitPython!=3.1.29,>=1.0.0 in /opt/conda/lib/python3.7/site-packages (from wandb) (3.1.30)
Requirement already satisfied: sentry-sdk>=1.0.0 in /opt/conda/lib/python3.7/site-packages (from wandb) (1.17.0)
Requirement already satisfied: psutil>=5.0.0 in /opt/conda/lib/python3.7/site-packages (from wandb) (5.9.3)
Requirement already satisfied: setproctitle in /opt/conda/lib/python3.7/site-packages (from wandb) (1.3.2)
Requirement already satisfied: docker-pycreds>=0.4.0 in /opt/conda/lib/python3.7/site-packages (from wandb) (0.4.0)
Requirement already satisfied: Click!=8.0.0,>=7.0 in /opt/conda/lib/python3.7/site-packages (from wandb) (8.1.3)
Requirement already satisfied: requests<3,>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from wandb) (2.28.2)
Requirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from Click!=8.0.0,>=7.0->wandb) (4.11.4)
Requirement already satisfied: six>=1.4.0 in /opt/conda/lib/python3.7/site-packages (from docker-pycreds>=0.4.0->wandb) (1.16.0)
Requirement already satisfied: gitdb<5,>=4.0.1 in /opt/conda/lib/python3.7/site-packages (from GitPython!=3.1.29,>=1.0.0->wandb) (4.0.10)
Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.0.0->wandb) (2022.12.7)
Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.0.0->wandb) (3.4)
Requirement already satisfied: charset-normalizer<4,>=2 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.0.0->wandb) (2.1.1)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.0.0->wandb) (1.26.14)
Requirement already satisfied: smmap<6,>=3.0.1 in /opt/conda/lib/python3.7/site-packages (from gitdb<5,>=4.0.1->GitPython!=3.1.29,>=1.0.0->wandb) (5.0.0)
Requirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->Click!=8.0.0,>=7.0->wandb) (3.11.0)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: pandas in /opt/conda/lib/python3.7/site-packages (1.3.5)
Requirement already satisfied: python-dateutil>=2.7.3 in /opt/conda/lib/python3.7/site-packages (from pandas) (2.8.2)
Requirement already satisfied: pytz>=2017.3 in /opt/conda/lib/python3.7/site-packages (from pandas) (2023.2)
Requirement already satisfied: numpy>=1.17.3 in /opt/conda/lib/python3.7/site-packages (from pandas) (1.21.6)
Requirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.7.3->pandas) (1.16.0)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: matplotlib in /opt/conda/lib/python3.7/site-packages (3.5.3)
Requirement already satisfied: numpy>=1.17 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (1.21.6)
Requirement already satisfied: fonttools>=4.22.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (4.38.0)
Requirement already satisfied: cycler>=0.10 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (0.11.0)
Requirement already satisfied: python-dateutil>=2.7 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (2.8.2)
Requirement already satisfied: packaging>=20.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (23.0)
Requirement already satisfied: pyparsing>=2.2.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (3.0.9)
Requirement already satisfied: pillow>=6.2.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (9.4.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib) (1.4.4)
Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from kiwisolver>=1.0.1->matplotlib) (4.4.0)
Requirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.7->matplotlib) (1.16.0)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
In [7]:
!pip install -qr requirements.txt

WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
In [8]:
import yolov5

Import the necessary libraries Suppress warning messages that may occur during the execution of a Python program

In [9]:
import pandas as pd
import numpy as np
import os
import glob
from datetime import datetime
import xml.etree.ElementTree as ET 
import cv2
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')

The path of plate annotations and car images

In [10]:
a_path = "/kaggle/input/car-plate-detection/annotations"
i_path = "/kaggle/input/car-plate-detection/images"

Set the list for img and plate location storage

In [11]:
X = []
y = []

Build new dataset to store the annotation information

In [12]:
dataset = {
            "filename":[],   
            "width":[],
            "height":[],
            "xmin":[],
            "ymin":[],   
            "xmax":[],
            "ymax":[],
           }
In [13]:
from lxml import etree

Extracting information from XML files that contain annotations for images, such as image size (width and height) and bounding box coordinates (xmin, ymin, xmax, ymax) for object detection tasks.

In [14]:
# extract the size of image
for annotation in glob.glob(a_path+"/*.xml"): #extract xml files
    ETtree = etree.parse(annotation)
    tree = ET.parse(annotation)
    
    # according to tag in xml file, extract the information of img and plate location
    for dim in ETtree.xpath("size"):
        width = int(dim.xpath("width")[0].text)
        height = int(dim.xpath("height")[0].text)
        
        
    for dim in ETtree.xpath("object/bndbox"):
        xmin = int(dim.xpath("xmin")[0].text)
        ymin = int(dim.xpath("ymin")[0].text)
        xmax = int(dim.xpath("xmax")[0].text)
        ymax = int(dim.xpath("ymax")[0].text)
        y.append([int(xmax), int(ymax), int(xmin), int(ymin)]) # put the plate location information in a list
        
    #get the name of the xml file and find the related img (some pictures have multiple plates)
        data_name=[annotation.split('/')[-1][0:-4]] 
        img_name = data_name[0]
        img_name = img_name + ".png"
        img_path = os.path.join(i_path, img_name)
        img_path = img_path.replace("\\", "/")
        img1 = glob.glob(img_path)
        img = cv2.imread(img1[0])
        X.append(np.array(img))
        
        dataset['filename'].append(img_name)
        dataset['width'].append(width)
        dataset['height'].append(height)
        dataset['xmin'].append(xmin)
        dataset['ymin'].append(ymin)
        dataset['xmax'].append(xmax)
        dataset['ymax'].append(ymax)
In [15]:
data_plate=pd.DataFrame(dataset)
data_plate
Out[15]:
filename width height xmin ymin xmax ymax
0 Cars339.png 500 300 209 135 283 169
1 Cars13.png 400 268 191 147 242 169
2 Cars74.png 400 267 115 115 277 153
3 Cars16.png 400 221 36 175 62 186
4 Cars291.png 517 303 71 205 215 246
... ... ... ... ... ... ... ...
466 Cars166.png 300 400 148 123 201 150
467 Cars60.png 400 300 45 98 364 159
468 Cars52.png 400 300 226 181 327 210
469 Cars297.png 400 233 158 149 247 170
470 Cars349.png 400 267 38 225 146 249

471 rows × 7 columns

In [16]:
X[2]
Out[16]:
array([[[ 29,  72,  67],
        [ 53,  94,  92],
        [ 63, 104, 102],
        ...,
        [117, 104,  84],
        [118, 105,  85],
        [121, 107,  89]],

       [[ 31,  74,  69],
        [ 53,  95,  92],
        [ 63, 103, 102],
        ...,
        [120, 108,  87],
        [120, 107,  87],
        [123, 110,  91]],

       [[ 44,  90,  85],
        [ 56, 100,  96],
        [ 62, 105, 101],
        ...,
        [144, 132, 110],
        [138, 125, 103],
        [141, 128, 107]],

       ...,

       [[228, 227, 225],
        [214, 213, 211],
        [206, 206, 204],
        ...,
        [ 19,  18,  20],
        [ 10,   9,  11],
        [ 13,  12,  14]],

       [[203, 203, 199],
        [192, 191, 188],
        [197, 196, 193],
        ...,
        [ 27,  27,  28],
        [ 19,  17,  19],
        [ 14,  13,  14]],

       [[180, 182, 177],
        [178, 179, 175],
        [189, 188, 184],
        ...,
        [ 30,  30,  31],
        [ 16,  14,  16],
        [ 12,  11,  12]]], dtype=uint8)
In [17]:
y[2]
Out[17]:
[277, 153, 115, 115]

Display the image with the bounding box

In [18]:
image = cv2.rectangle(X[2],(y[2][0],y[2][1]),(y[2][2],y[2][3]),(0, 0, 255))
plt.imshow(image)
plt.show()
In [19]:
!mkdir "/kaggle/working/labels"

Processing image annotations in the YOLOv5 format. Normalizes the xmin, ymin, xmax, ymax values to YOLOv5 format by dividing them by the width and height of the image, adds the calculated x_center, y_center, frame_width, and frame_height values as new columns to the data_plate dataframe.

In [21]:
x_center = []
y_center = []
frame_width = []
frame_height = []

save_type = 'w'

for i, row in enumerate(data_plate.iloc):
    # delete the ".png" in filename
    current_filename = str(row.filename[:-4])
    
    # extract the information of image
    width, height, xmin, ymin, xmax, ymax = list(data_plate.iloc[i][-6:])
    
    # normalize the information to YOLOV5 type
    x=(xmin+xmax)/2/width
    y=(ymin+ymax)/2/height
    width=(xmax-xmin)/width
    height=(ymax-ymin)/height
    
    x_center.append(x)
    y_center.append(y)
    frame_width.append(width)
    frame_height.append(height)
    
    # create a string with the YOLOv5 format label
    txt = '0' + ' ' + str(x) + ' ' + str(y) + ' ' + str(width) + ' ' + str(height) + '\n'
    
    # create separate annotation files for each image
    if i > 0:
        previous_filename = str(data_plate.filename[i-1][:-4])
        save_type='a+' if current_filename == previous_filename else 'w'
    
    with open("/kaggle/working/labels/" + str(row.filename[:-4]) +'.txt', save_type) as f:
        f.write(txt)

        
data_plate['x_center']=x_center
data_plate['y_center']=y_center
data_plate['frame_width']=frame_width
data_plate['frame_height']=frame_height
In [22]:
import os

files = os.listdir('/kaggle/working/labels')
print(files[0])

Cars145.txt
In [23]:
len(files)
Out[23]:
433
In [24]:
data_plate
Out[24]:
filename width height xmin ymin xmax ymax x_center y_center frame_width frame_height
0 Cars339.png 500 300 209 135 283 169 0.492000 0.506667 0.148000 0.113333
1 Cars13.png 400 268 191 147 242 169 0.541250 0.589552 0.127500 0.082090
2 Cars74.png 400 267 115 115 277 153 0.490000 0.501873 0.405000 0.142322
3 Cars16.png 400 221 36 175 62 186 0.122500 0.816742 0.065000 0.049774
4 Cars291.png 517 303 71 205 215 246 0.276596 0.744224 0.278530 0.135314
... ... ... ... ... ... ... ... ... ... ... ...
466 Cars166.png 300 400 148 123 201 150 0.581667 0.341250 0.176667 0.067500
467 Cars60.png 400 300 45 98 364 159 0.511250 0.428333 0.797500 0.203333
468 Cars52.png 400 300 226 181 327 210 0.691250 0.651667 0.252500 0.096667
469 Cars297.png 400 233 158 149 247 170 0.506250 0.684549 0.222500 0.090129
470 Cars349.png 400 267 38 225 146 249 0.230000 0.887640 0.270000 0.089888

471 rows × 11 columns

In [25]:
duplicate_filenames = data_plate[data_plate['filename'].duplicated(keep=False)]['filename'].unique()
duplicate_rows = data_plate[data_plate['filename'].isin(duplicate_filenames)]
print(duplicate_rows)

        filename  width  height  xmin  ymin  xmax  ymax  x_center  y_center  \
7    Cars132.png    400     225    23   190    56   198   0.09875  0.862222   
8    Cars132.png    400     225   378   188   400   200   0.97250  0.862222   
48   Cars295.png    400     256    52   170    73   182   0.15625  0.687500   
49   Cars295.png    400     256   237   143   271   162   0.63500  0.595703   
81   Cars358.png    400     200    44    94   107   107   0.18875  0.502500   
..           ...    ...     ...   ...   ...   ...   ...       ...       ...   
406  Cars413.png    400     256    51   170    73   183   0.15500  0.689453   
425  Cars103.png    400     196   230   129   248   134   0.59750  0.670918   
426  Cars103.png    400     196   189   116   202   121   0.48875  0.604592   
448   Cars85.png    400     267    86   218   146   235   0.29000  0.848315   
449   Cars85.png    400     267   313   129   358   141   0.83875  0.505618   

     frame_width  frame_height  
7         0.0825      0.035556  
8         0.0550      0.053333  
48        0.0525      0.046875  
49        0.0850      0.074219  
81        0.1575      0.065000  
..           ...           ...  
406       0.0550      0.050781  
425       0.0450      0.025510  
426       0.0325      0.025510  
448       0.1500      0.063670  
449       0.1125      0.044944  

[62 rows x 11 columns]
In [26]:
# some images have multiple bounding box with separate information
filtered_data = data_plate[data_plate['filename'] == 'Cars295.png']
filtered_data
Out[26]:
filename width height xmin ymin xmax ymax x_center y_center frame_width frame_height
48 Cars295.png 400 256 52 170 73 182 0.15625 0.687500 0.0525 0.046875
49 Cars295.png 400 256 237 143 271 162 0.63500 0.595703 0.0850 0.074219

Images with multiple bboxes store consistent yolov5 information in one txt file

In [27]:
with open('/kaggle/working/labels/Cars295.txt', 'r') as file:
    contents = file.read()
    print(contents)

0 0.15625 0.6875 0.0525 0.046875
0 0.635 0.595703125 0.085 0.07421875
In [28]:
# create a list of img names
img_names = [*os.listdir("/kaggle/input/car-plate-detection/images")]
In [29]:
img_names[0:5]
Out[29]:
['Cars393.png', 'Cars376.png', 'Cars87.png', 'Cars190.png', 'Cars177.png']
In [30]:
data_plate['filename'] = data_plate['filename'].str.replace('.png', '')
In [31]:
data_plate
Out[31]:
filename width height xmin ymin xmax ymax x_center y_center frame_width frame_height
0 Cars339 500 300 209 135 283 169 0.492000 0.506667 0.148000 0.113333
1 Cars13 400 268 191 147 242 169 0.541250 0.589552 0.127500 0.082090
2 Cars74 400 267 115 115 277 153 0.490000 0.501873 0.405000 0.142322
3 Cars16 400 221 36 175 62 186 0.122500 0.816742 0.065000 0.049774
4 Cars291 517 303 71 205 215 246 0.276596 0.744224 0.278530 0.135314
... ... ... ... ... ... ... ... ... ... ... ...
466 Cars166 300 400 148 123 201 150 0.581667 0.341250 0.176667 0.067500
467 Cars60 400 300 45 98 364 159 0.511250 0.428333 0.797500 0.203333
468 Cars52 400 300 226 181 327 210 0.691250 0.651667 0.252500 0.096667
469 Cars297 400 233 158 149 247 170 0.506250 0.684549 0.222500 0.090129
470 Cars349 400 267 38 225 146 249 0.230000 0.887640 0.270000 0.089888

471 rows × 11 columns

In [32]:
# split the images into training, validation and testing sets
from sklearn.model_selection import train_test_split
train, test = train_test_split(img_names, test_size=0.2, random_state=42)
test, val = train_test_split(test, test_size=0.7, random_state=42)
In [33]:
# create train, validation and test dir
os.chdir('/kaggle/working/')
os.mkdir('./yolov5/data/train')
os.mkdir('./yolov5/data/val')
os.mkdir('./yolov5/data/test')
os.mkdir('./yolov5/data/train/images')
os.mkdir('./yolov5/data/train/labels')
os.mkdir('./yolov5/data/test/images')
os.mkdir('./yolov5/data/test/labels')
os.mkdir('./yolov5/data/val/images')
os.mkdir('./yolov5/data/val/labels')

The function copyImages takes a list of image filenames (imageList) and a folder name (folder_Name) as input. Uses the PIL library to open each image from the input folder, and then saves the image to a new location in the YOLOv5 format,

In [34]:
# define the images copy function
from PIL import Image

def copyImages(imageList, folder_Name):
    for image in imageList:
        img = Image.open("../input/car-plate-detection/images/"+image)
        img.save("./yolov5/data/"+folder_Name+"/images/"+image)
In [35]:
copyImages(train, "train")
copyImages(val, "val")
copyImages(test, "test")

Check if the images have been splited to target folder

In [36]:
path = "/kaggle/working/yolov5/data/train/images"

files = os.listdir(path)

print(files[0])
print(len(files))

Cars139.png
346
In [37]:
data_string = data_plate.astype('string')

The function writes the generated string to a .txt file with the same name as the image, but with a .txt extension, in the directory "./yolov5/data/"+data_name+"/labels/". This file will contain the labels for the corresponding image in the YOLOv5 format.

In [38]:
def create_labels(image_list, data_name):
    #create a list of image names without .img
    fileNames = [x.split(".")[0] for x in image_list]
    
    # create dataframe where 'filename' column = name
    for name in fileNames:
        data = data_string[data_string.filename==name]
        box_list = []
        
        # add object class and 4 box parameters in box_list
        for index in range(len(data)):
            row = data.iloc[index]
            box_list.append(str(0)+" "+row["x_center"]+" "+row["y_center"]\
                        +" "+row["frame_width"]+" "+row["frame_height"])
        
        # convert the bound box information to a string for YOLOV5 to detect
        text = "\n".join(box_list)
        with open("./yolov5/data/"+data_name+"/labels/"+name+".txt", "w") as file:
            file.write(text)
In [39]:
create_labels(train, "train")
create_labels(val, "val")
create_labels(test, "test")

Check if the label files have been created

In [40]:
folder_path = './yolov5/data/train/labels'
files = os.listdir(folder_path)

for file in files:
    print(file)
    break

Cars300.txt
In [41]:
with open('/kaggle/working/yolov5/data/train/labels/Cars295.txt', 'r') as file:
    contents = file.read()
    print(contents)

0 0.15625 0.6875 0.0525 0.046875
0 0.635 0.595703125 0.085 0.07421875
In [42]:
# go to YOLO directory
%cd yolov5

/kaggle/working/yolov5

Initializes Notebook display to show images produced by the YOLOv5 model.

In [43]:
from IPython.display import Image, clear_output
import torch
from yolov5 import utils
display = utils.notebook_init()

YOLOv5 🚀 v7.0-133-gcca5e21 Python-3.7.12 torch-1.9.0+cu111 CUDA:0 (Tesla P100-PCIE-16GB, 16281MiB)

Setup complete ✅ (2 CPUs, 15.6 GB RAM, 4557.6/8062.4 GB disk)
In [44]:
# define the train and val directories of yaml text, nc means number of class and names is the class name
yaml_text = """train: data/train/images
val: data/val/images

nc: 1
names: ['license']"""
In [45]:
# write yaml text to data.yaml
with open("data/data.yaml", 'w') as file:
    file.write(yaml_text)
In [46]:
with open("data/data.yaml", 'r') as file:
    print(file.read())

train: data/train/images
val: data/val/images

nc: 1
names: ['license']

The %writetemplate magic command takes a filename as the line argument and a cell as the cell argument. It opens the file in write mode using the filename provided, and writes the contents of the cell to the file using the format() method to replace placeholders with the values of variables from the global namespace using **globals().

In [47]:
#customize iPython writefile to write variables
from IPython.core.magic import register_line_cell_magic

@register_line_cell_magic
def writetemplate(line, cell):
    with open(line, 'w') as f:
        f.write(cell.format(**globals()))

This model has a YOLOv5s backbone with a custom head, designed for one-class object detection tasks. The backbone consists of a series of convolutional layers and bottleneck blocks, which are used to extract features from the input image.

In [48]:
%%writetemplate models/custom_yolov5s.yaml

# parameters
nc: 1  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.50  # layer channel multiple

# anchors
anchors:
    - [10,13, 16,30, 33,23]  
    - [30,61, 62,45, 59,119]  
    - [116,90, 156,198, 373,326] 
    
# YOLOv5 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Focus, [64, 3]], 
   [-1, 1, Conv, [128, 3, 2]],  
   [-1, 3, BottleneckCSP, [128]],
   [-1, 1, Conv, [256, 3, 2]],  
   [-1, 9, BottleneckCSP, [256]],
   [-1, 1, Conv, [512, 3, 2]],  
   [-1, 9, BottleneckCSP, [512]],
   [-1, 1, Conv, [1024, 3, 2]],  
   [-1, 1, SPP, [1024, [5, 9, 13]]],
   [-1, 3, BottleneckCSP, [1024, False]],  
  ]

# YOLOv5 head
head:
    [[-1, 1, Conv, [512, 1, 1]],
    [-1, 1, nn.Upsample, [None, 2, 'nearest']],
    [[-1, 6], 1, Concat, [1]],  
    [-1, 3, BottleneckCSP, [512, False]],  

    [-1, 1, Conv, [256, 1, 1]],
    [-1, 1, nn.Upsample, [None, 2, 'nearest']],
    [[-1, 4], 1, Concat, [1]],  
    [-1, 3, BottleneckCSP, [256, False]],  

    [-1, 1, Conv, [256, 3, 2]],
    [[-1, 14], 1, Concat, [1]],  
    [-1, 3, BottleneckCSP, [512, False]],  

    [-1, 1, Conv, [512, 3, 2]],
    [[-1, 10], 1, Concat, [1]],  
    [-1, 3, BottleneckCSP, [1024, False]], 

    [[17, 20, 23], 1, Detect, [nc, anchors]],  
    ]
In [49]:
# train yolov5s on data
# record the time of training
# --img: size of image; 
# --data: The path to the data configuration file
# --cfg: The path to the model configuration file
# --name: The name of the results folder
# --cache: Enable caching images for faster training

start = datetime.now()
!python train.py --img 640 --batch 32 --epochs 50 --data data/data.yaml --cfg models/custom_yolov5s.yaml --weights yolov5s.pt --name yolov5s_results  --cache
end = datetime.now()

wandb: WARNING ⚠️ wandb is deprecated and will be removed in a future release. See supported integrations at https://github.com/ultralytics/yolov5#integrations.
wandb: (1) Create a W&B account
wandb: (2) Use an existing W&B account
wandb: (3) Don't visualize my results
wandb: Enter your choice: (30 second timeout) 
wandb: W&B disabled due to login timeout.
train: weights=yolov5s.pt, cfg=models/custom_yolov5s.yaml, data=data/data.yaml, hyp=data/hyps/hyp.scratch-low.yaml, epochs=50, batch_size=32, imgsz=640, rect=False, resume=False, nosave=False, noval=False, noautoanchor=False, noplots=False, evolve=None, bucket=, cache=ram, image_weights=False, device=, multi_scale=False, single_cls=False, optimizer=SGD, sync_bn=False, workers=8, project=runs/train, name=yolov5s_results, exist_ok=False, quad=False, cos_lr=False, label_smoothing=0.0, patience=100, freeze=[0], save_period=-1, seed=0, local_rank=-1, entity=None, upload_dataset=False, bbox_interval=-1, artifact_alias=latest
github: up to date with https://github.com/ultralytics/yolov5 ✅
requirements: YOLOv5 requirements "tqdm>=4.64.0" "tensorboard>=2.4.1" not found, attempting AutoUpdate...
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: tqdm>=4.64.0 in /opt/conda/lib/python3.7/site-packages (4.64.1)
Requirement already satisfied: tensorboard>=2.4.1 in /opt/conda/lib/python3.7/site-packages (2.11.2)
Requirement already satisfied: protobuf<4,>=3.9.2 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (3.20.3)
Requirement already satisfied: setuptools>=41.0.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (67.6.1)
Requirement already satisfied: grpcio>=1.24.3 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (1.51.1)
Requirement already satisfied: werkzeug>=1.0.1 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (2.2.3)
Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (0.6.1)
Requirement already satisfied: wheel>=0.26 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (0.38.4)
Requirement already satisfied: markdown>=2.6.8 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (3.4.1)
Requirement already satisfied: requests<3,>=2.21.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (2.28.2)
Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (0.4.6)
Requirement already satisfied: absl-py>=0.4 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (1.4.0)
Requirement already satisfied: numpy>=1.12.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (1.21.6)
Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (1.8.1)
Requirement already satisfied: google-auth<3,>=1.6.3 in /opt/conda/lib/python3.7/site-packages (from tensorboard>=2.4.1) (1.35.0)
Requirement already satisfied: six>=1.9.0 in /opt/conda/lib/python3.7/site-packages (from google-auth<3,>=1.6.3->tensorboard>=2.4.1) (1.16.0)
Requirement already satisfied: pyasn1-modules>=0.2.1 in /opt/conda/lib/python3.7/site-packages (from google-auth<3,>=1.6.3->tensorboard>=2.4.1) (0.2.8)
Requirement already satisfied: rsa<5,>=3.1.4 in /opt/conda/lib/python3.7/site-packages (from google-auth<3,>=1.6.3->tensorboard>=2.4.1) (4.9)
Requirement already satisfied: cachetools<5.0,>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from google-auth<3,>=1.6.3->tensorboard>=2.4.1) (4.2.4)
Requirement already satisfied: requests-oauthlib>=0.7.0 in /opt/conda/lib/python3.7/site-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard>=2.4.1) (1.3.1)
Requirement already satisfied: importlib-metadata>=4.4 in /opt/conda/lib/python3.7/site-packages (from markdown>=2.6.8->tensorboard>=2.4.1) (4.11.4)
Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard>=2.4.1) (3.4)
Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard>=2.4.1) (2022.12.7)
Requirement already satisfied: charset-normalizer<4,>=2 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard>=2.4.1) (2.1.1)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard>=2.4.1) (1.26.14)
Requirement already satisfied: MarkupSafe>=2.1.1 in /opt/conda/lib/python3.7/site-packages (from werkzeug>=1.0.1->tensorboard>=2.4.1) (2.1.1)
Requirement already satisfied: typing-extensions>=3.6.4 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard>=2.4.1) (4.4.0)
Requirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard>=2.4.1) (3.11.0)
Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /opt/conda/lib/python3.7/site-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard>=2.4.1) (0.4.8)
Requirement already satisfied: oauthlib>=3.0.0 in /opt/conda/lib/python3.7/site-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard>=2.4.1) (3.2.2)

requirements: 2 packages updated per /kaggle/working/yolov5/requirements.txt
requirements: ⚠️ Restart runtime or rerun command for updates to take effect

YOLOv5 🚀 v7.0-133-gcca5e21 Python-3.7.12 torch-1.9.0+cu111 CUDA:0 (Tesla P100-PCIE-16GB, 16281MiB)

hyperparameters: lr0=0.01, lrf=0.01, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0, copy_paste=0.0
ClearML: run 'pip install clearml' to automatically track, visualize and remotely train YOLOv5 🚀 in ClearML
Comet: run 'pip install comet_ml' to automatically track and visualize YOLOv5 🚀 runs in Comet
TensorBoard: Start with 'tensorboard --logdir runs/train', view at http://localhost:6006/
Downloading https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s.pt to yolov5s.pt...
100%|██████████████████████████████████████| 14.1M/14.1M [00:00<00:00, 77.5MB/s]


                 from  n    params  module                                  arguments                     
  0                -1  1      3520  models.common.Focus                     [3, 32, 3]                    
  1                -1  1     18560  models.common.Conv                      [32, 64, 3, 2]                
  2                -1  1     19904  models.common.BottleneckCSP             [64, 64, 1]                   
  3                -1  1     73984  models.common.Conv                      [64, 128, 3, 2]               
  4                -1  3    161152  models.common.BottleneckCSP             [128, 128, 3]                 
  5                -1  1    295424  models.common.Conv                      [128, 256, 3, 2]              
  6                -1  3    641792  models.common.BottleneckCSP             [256, 256, 3]                 
  7                -1  1   1180672  models.common.Conv                      [256, 512, 3, 2]              
  8                -1  1    656896  models.common.SPP                       [512, 512, [5, 9, 13]]        
  9                -1  1   1248768  models.common.BottleneckCSP             [512, 512, 1, False]          
 10                -1  1    131584  models.common.Conv                      [512, 256, 1, 1]              
 11                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 12           [-1, 6]  1         0  models.common.Concat                    [1]                           
 13                -1  1    378624  models.common.BottleneckCSP             [512, 256, 1, False]          
 14                -1  1     33024  models.common.Conv                      [256, 128, 1, 1]              
 15                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 16           [-1, 4]  1         0  models.common.Concat                    [1]                           
 17                -1  1     95104  models.common.BottleneckCSP             [256, 128, 1, False]          
 18                -1  1    147712  models.common.Conv                      [128, 128, 3, 2]              
 19          [-1, 14]  1         0  models.common.Concat                    [1]                           
 20                -1  1    313088  models.common.BottleneckCSP             [256, 256, 1, False]          
 21                -1  1    590336  models.common.Conv                      [256, 256, 3, 2]              
 22          [-1, 10]  1         0  models.common.Concat                    [1]                           
 23                -1  1   1248768  models.common.BottleneckCSP             [512, 512, 1, False]          
 24      [17, 20, 23]  1     16182  models.yolo.Detect                      [1, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]], [128, 256, 512]]
custom_YOLOv5s summary: 233 layers, 7255094 parameters, 7255094 gradients

Transferred 223/369 items from yolov5s.pt
AMP: checks passed ✅
optimizer: SGD(lr=0.01) with parameter groups 59 weight(decay=0.0), 70 weight(decay=0.0005), 62 bias
albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01), CLAHE(p=0.01, clip_limit=(1, 4.0), tile_grid_size=(8, 8))
train: Scanning /kaggle/working/yolov5/data/train/labels... 346 images, 0 backgr
train: New cache created: /kaggle/working/yolov5/data/train/labels.cache
train: Caching images (0.3GB ram): 100%|██████████| 346/346 [00:02<00:00, 156.71
val: Scanning /kaggle/working/yolov5/data/val/labels... 61 images, 0 backgrounds
val: New cache created: /kaggle/working/yolov5/data/val/labels.cache
val: Caching images (0.0GB ram): 100%|██████████| 61/61 [00:00<00:00, 65.73it/s]
[W pthreadpool-cpp.cc:90] Warning: Leaking Caffe2 thread-pool after fork. (function pthreadpool)
[W pthreadpool-cpp.cc:90] Warning: Leaking Caffe2 thread-pool after fork. (function pthreadpool)

AutoAnchor: 4.11 anchors/target, 1.000 Best Possible Recall (BPR). Current anchors are a good fit to dataset ✅
Plotting labels to runs/train/yolov5s_results/labels.jpg... 
Image sizes 640 train, 640 val
Using 2 dataloader workers
Logging results to runs/train/yolov5s_results
Starting training for 50 epochs...

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       0/49      7.41G     0.1115    0.02685          0         60        640:  train.py:323: FutureWarning: Non-finite norm encountered in torch.nn.utils.clip_grad_norm_; continuing anyway. Note that the default behavior will change in a future release to error out if a non-finite total norm is encountered. At that point, setting error_if_nonfinite=false will be required to retain the old behavior.
  torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=10.0)  # clip gradients
       0/49      8.83G     0.1106    0.02752          0         58        640: 1
                 Class     Images  Instances          P          R      mAP50   WARNING ⚠️ NMS time limit 3.550s exceeded
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69   0.000123     0.0145   6.59e-05   2.64e-05

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       1/49      8.83G     0.1048    0.02641          0         54        640: 1
                 Class     Images  Instances          P          R      mAP50   WARNING ⚠️ NMS time limit 3.550s exceeded
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69   0.000115     0.0145   5.96e-05   2.98e-05

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       2/49      8.83G     0.1006    0.02568          0         82        640:  train.py:323: FutureWarning: Non-finite norm encountered in torch.nn.utils.clip_grad_norm_; continuing anyway. Note that the default behavior will change in a future release to error out if a non-finite total norm is encountered. At that point, setting error_if_nonfinite=false will be required to retain the old behavior.
  torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=10.0)  # clip gradients
       2/49      8.83G     0.1001    0.02529          0         47        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69   0.000656      0.174   0.000398   0.000112

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       3/49      8.83G    0.09659      0.025          0         52        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69   0.000437      0.116   0.000264    8.2e-05

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       4/49      8.83G    0.09343    0.02416          0         54        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69   0.000219      0.058   0.000121   3.26e-05

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       5/49      8.83G    0.08991    0.02412          0         54        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69   0.000546      0.145   0.000375    0.00012

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       6/49      8.83G    0.08667     0.0244          0         47        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69   0.000546      0.145   0.000403    0.00014

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       7/49      8.83G    0.08328    0.02387          0         67        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69    0.00109       0.29    0.00388   0.000723

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       8/49      8.83G    0.07852    0.02473          0         60        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69    0.00082      0.217    0.00413    0.00106

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
       9/49      8.83G    0.07636    0.02407          0         62        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69    0.00251      0.667    0.00485    0.00128

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      10/49      8.83G     0.0726    0.02416          0         65        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69    0.00257      0.681     0.0626      0.013

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      11/49      8.83G    0.07056    0.02373          0         67        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69     0.0601      0.188     0.0577     0.0123

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      12/49      8.83G    0.06751    0.02166          0         41        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.258       0.42      0.281     0.0694

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      13/49      8.83G    0.06459    0.02099          0         45        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.184      0.261       0.13     0.0309

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      14/49      8.83G    0.06164    0.02086          0         57        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.376      0.391      0.301     0.0962

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      15/49      8.83G    0.06065    0.02002          0         50        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69       0.39      0.348      0.284     0.0776

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      16/49      8.83G    0.06014     0.0201          0         60        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.464      0.319      0.272     0.0709

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      17/49      8.83G    0.05726    0.01888          0         50        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.413      0.377      0.341      0.127

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      18/49      8.83G     0.0564    0.01893          0         51        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.628      0.493      0.472      0.142

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      19/49      8.83G     0.0553    0.01922          0         49        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.847      0.449      0.567      0.208

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      20/49      8.83G    0.05325     0.0172          0         52        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.658      0.551      0.535       0.18

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      21/49      8.83G    0.05227    0.01784          0         44        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.755      0.446      0.468      0.206

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      22/49      8.83G    0.05051    0.01607          0         41        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.506      0.464      0.474      0.153

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      23/49      8.83G    0.05168    0.01764          0         57        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.834      0.522       0.59      0.199

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      24/49      8.83G    0.04976    0.01539          0         51        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.798      0.515      0.568      0.256

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      25/49      8.83G    0.04939    0.01552          0         47        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.853      0.591      0.656      0.226

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      26/49      8.83G    0.04906    0.01532          0         50        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.849       0.58      0.621      0.232

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      27/49      8.83G    0.04682    0.01521          0         57        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.722      0.609      0.602      0.249

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      28/49      8.83G    0.04646    0.01461          0         60        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.885      0.536      0.643      0.271

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      29/49      8.83G     0.0459    0.01556          0         56        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.778       0.56      0.674      0.285

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      30/49      8.83G    0.04532     0.0139          0         45        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.889      0.565      0.709      0.346

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      31/49      8.83G    0.04387    0.01349          0         58        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.925      0.565      0.709      0.322

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      32/49      8.83G    0.04335    0.01399          0         56        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.764      0.652      0.717       0.33

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      33/49      8.83G    0.04237    0.01363          0         55        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.797      0.609      0.704      0.327

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      34/49      8.83G    0.04264    0.01417          0         53        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.929      0.609      0.719       0.33

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      35/49      8.83G      0.042    0.01411          0         60        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.946      0.594      0.718      0.346

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      36/49      8.83G    0.04062    0.01341          0         41        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.977      0.605      0.758      0.349

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      37/49      8.83G    0.04016    0.01266          0         55        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69       0.94      0.623      0.765      0.374

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      38/49      8.83G    0.03911    0.01385          0         50        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.832      0.652      0.737      0.364

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      39/49      8.83G    0.04006    0.01272          0         47        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.787      0.695      0.772       0.39

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      40/49      8.83G    0.03915    0.01331          0         59        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.812      0.681      0.787      0.379

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      41/49      8.83G    0.03785    0.01297          0         59        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.955      0.609      0.801      0.413

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      42/49      8.83G     0.0376     0.0121          0         50        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.702       0.71      0.744      0.382

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      43/49      8.83G    0.03681    0.01221          0         50        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.762      0.754      0.777        0.4

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      44/49      8.83G    0.03758    0.01268          0         66        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.832      0.739      0.778      0.402

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      45/49      8.83G    0.03554    0.01176          0         51        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69       0.81      0.741      0.786        0.4

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      46/49      8.83G    0.03598    0.01197          0         38        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.821      0.731      0.805      0.409

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      47/49      8.83G     0.0357    0.01168          0         55        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.859      0.739      0.819      0.427

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      48/49      8.83G    0.03593    0.01212          0         59        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.845      0.739      0.817      0.419

      Epoch    GPU_mem   box_loss   obj_loss   cls_loss  Instances       Size
      49/49      8.83G    0.03504     0.0121          0         48        640: 1
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69      0.823       0.71      0.812      0.434

50 epochs completed in 0.091 hours.
Optimizer stripped from runs/train/yolov5s_results/weights/last.pt, 14.9MB
Optimizer stripped from runs/train/yolov5s_results/weights/best.pt, 14.9MB

Validating runs/train/yolov5s_results/weights/best.pt...
Fusing layers... 
custom_YOLOv5s summary: 182 layers, 7246518 parameters, 0 gradients
                 Class     Images  Instances          P          R      mAP50   
                   all         61         69       0.82       0.71      0.812      0.432
Results saved to runs/train/yolov5s_results
In [50]:
print("Runtime =",end-start)

Runtime = 0:06:48.681023
In [51]:
# check the train model result list
file_list = os.listdir('runs/train/yolov5s_results')
file_list
Out[51]:
['val_batch0_labels.jpg',
 'confusion_matrix.png',
 'results.csv',
 'labels.jpg',
 'opt.yaml',
 'F1_curve.png',
 'train_batch1.jpg',
 'labels_correlogram.jpg',
 'train_batch0.jpg',
 'train_batch2.jpg',
 'events.out.tfevents.1680435340.3596ba3427d1.1223.0',
 'PR_curve.png',
 'hyp.yaml',
 'results.png',
 'weights',
 'val_batch0_pred.jpg',
 'P_curve.png',
 'R_curve.png']
In [52]:
plt.figure(figsize=(30,15))
plt.axis('off')
plt.imshow(plt.imread('/kaggle/working/yolov5/runs/train/yolov5s_results/results.png'))
Out[52]:
<matplotlib.image.AxesImage at 0x7f7b45d60f90>
In [53]:
# visualize the train results
img = plt.imread('runs/train/yolov5s_results/train_batch2.jpg')
plt.figure(figsize=(20,15))
plt.imshow(img)
plt.axis('off')
plt.show()
In [54]:
# apply YOLOv5 framework on test images
# --conf 0.4: 
!python detect.py --source data/test/images/ --weight runs/train/yolov5s_results/weights/best.pt --name expTestImage --conf 0.4

detect: weights=['runs/train/yolov5s_results/weights/best.pt'], source=data/test/images/, data=data/coco128.yaml, imgsz=[640, 640], conf_thres=0.4, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=expTestImage, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False, vid_stride=1
requirements: YOLOv5 requirement "tqdm>=4.64.0" not found, attempting AutoUpdate...
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: tqdm>=4.64.0 in /opt/conda/lib/python3.7/site-packages (4.64.1)

requirements: 1 package updated per /kaggle/working/yolov5/requirements.txt
requirements: ⚠️ Restart runtime or rerun command for updates to take effect

YOLOv5 🚀 v7.0-133-gcca5e21 Python-3.7.12 torch-1.9.0+cu111 CUDA:0 (Tesla P100-PCIE-16GB, 16281MiB)

Fusing layers... 
custom_YOLOv5s summary: 182 layers, 7246518 parameters, 0 gradients
image 1/26 /kaggle/working/yolov5/data/test/images/Cars100.png: 448x640 1 license, 10.5ms
image 2/26 /kaggle/working/yolov5/data/test/images/Cars102.png: 512x640 (no detections), 12.2ms
image 3/26 /kaggle/working/yolov5/data/test/images/Cars105.png: 384x640 3 licenses, 10.6ms
image 4/26 /kaggle/working/yolov5/data/test/images/Cars110.png: 448x640 1 license, 7.7ms
image 5/26 /kaggle/working/yolov5/data/test/images/Cars145.png: 448x640 2 licenses, 7.3ms
image 6/26 /kaggle/working/yolov5/data/test/images/Cars151.png: 416x640 1 license, 11.0ms
image 7/26 /kaggle/working/yolov5/data/test/images/Cars153.png: 384x640 1 license, 7.9ms
image 8/26 /kaggle/working/yolov5/data/test/images/Cars171.png: 448x640 1 license, 7.9ms
image 9/26 /kaggle/working/yolov5/data/test/images/Cars210.png: 448x640 1 license, 7.4ms
image 10/26 /kaggle/working/yolov5/data/test/images/Cars230.png: 384x640 1 license, 9.7ms
image 11/26 /kaggle/working/yolov5/data/test/images/Cars243.png: 480x640 1 license, 11.7ms
image 12/26 /kaggle/working/yolov5/data/test/images/Cars246.png: 448x640 1 license, 7.9ms
image 13/26 /kaggle/working/yolov5/data/test/images/Cars26.png: 384x640 1 license, 8.1ms
image 14/26 /kaggle/working/yolov5/data/test/images/Cars263.png: 480x640 1 license, 7.5ms
image 15/26 /kaggle/working/yolov5/data/test/images/Cars265.png: 352x640 1 license, 11.4ms
image 16/26 /kaggle/working/yolov5/data/test/images/Cars277.png: 480x640 1 license, 7.7ms
image 17/26 /kaggle/working/yolov5/data/test/images/Cars278.png: 352x640 (no detections), 7.9ms
image 18/26 /kaggle/working/yolov5/data/test/images/Cars29.png: 480x640 1 license, 8.0ms
image 19/26 /kaggle/working/yolov5/data/test/images/Cars302.png: 320x640 2 licenses, 10.4ms
image 20/26 /kaggle/working/yolov5/data/test/images/Cars306.png: 640x640 1 license, 8.3ms
image 21/26 /kaggle/working/yolov5/data/test/images/Cars309.png: 384x640 1 license, 7.6ms
image 22/26 /kaggle/working/yolov5/data/test/images/Cars325.png: 384x640 1 license, 8.9ms
image 23/26 /kaggle/working/yolov5/data/test/images/Cars363.png: 448x640 1 license, 7.4ms
image 24/26 /kaggle/working/yolov5/data/test/images/Cars45.png: 416x640 1 license, 7.8ms
image 25/26 /kaggle/working/yolov5/data/test/images/Cars9.png: 512x640 1 license, 7.2ms
image 26/26 /kaggle/working/yolov5/data/test/images/Cars91.png: 416x640 1 license, 7.5ms
Speed: 0.4ms pre-process, 8.8ms inference, 0.8ms NMS per image at shape (1, 3, 640, 640)
Results saved to runs/detect/expTestImage
In [57]:
# show the trained test image example
dir_path = '/kaggle/working/yolov5/runs/detect/expTestImage'
files = os.listdir(dir_path)
files[1]
Out[57]:
'Cars246.png'
In [58]:
Image("/kaggle/working/yolov5/runs/detect/expTestImage/Cars246.png")
Out[58]:
In [59]:
folder_name = 'new_picture'

if not os.path.exists(folder_name):
    os.makedirs(folder_name)
In [61]:
# get some new images to test the YOLOV5 model
import shutil

shutil.copy('/kaggle/input/new-car-picture/new1.png', '/kaggle/working/yolov5/new_picture/new1.png')
shutil.copy('/kaggle/input/new-car-picture/new2.png', '/kaggle/working/yolov5/new_picture/new2.png')
Out[61]:
'/kaggle/working/yolov5/new_picture/new2.png'
In [62]:
dir_path = '/kaggle/working/yolov5/new_picture'
files = os.listdir(dir_path)
files
Out[62]:
['new1.png', 'new2.png']
In [63]:
!python detect.py --source "/kaggle/working/yolov5/new_picture" --weights '/kaggle/working/yolov5/runs/train/yolov5s_results/weights/best.pt'

detect: weights=['/kaggle/working/yolov5/runs/train/yolov5s_results/weights/best.pt'], source=/kaggle/working/yolov5/new_picture, data=data/coco128.yaml, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False, vid_stride=1
requirements: YOLOv5 requirement "tqdm>=4.64.0" not found, attempting AutoUpdate...
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: tqdm>=4.64.0 in /opt/conda/lib/python3.7/site-packages (4.64.1)

requirements: 1 package updated per /kaggle/working/yolov5/requirements.txt
requirements: ⚠️ Restart runtime or rerun command for updates to take effect

YOLOv5 🚀 v7.0-133-gcca5e21 Python-3.7.12 torch-1.9.0+cu111 CUDA:0 (Tesla P100-PCIE-16GB, 16281MiB)

Fusing layers... 
custom_YOLOv5s summary: 182 layers, 7246518 parameters, 0 gradients
image 1/2 /kaggle/working/yolov5/new_picture/new1.png: 288x640 1 license, 11.2ms
image 2/2 /kaggle/working/yolov5/new_picture/new2.png: 416x640 2 licenses, 10.2ms
Speed: 0.3ms pre-process, 10.7ms inference, 1.2ms NMS per image at shape (1, 3, 640, 640)
Results saved to runs/detect/exp
In [65]:
#show the results
Image("/kaggle/working/yolov5/runs/detect/exp/new2.png")
Out[65]:
In [ ]:
# https://www.kaggle.com/code/gowrishankarp/license-plate-detection-yolov5-pytesseract
# https://www.kaggle.com/code/saworz/realtime-plate-reading-video-yolov5-easyocr
# https://www.kaggle.com/code/rohitgadhwar/face-mask-detection-yolov5

Conclusion¶

The advantages of YOLOv5 include its speed, simplicity, and end-to-end training. It can be a good choice for real-time object detection tasks, such as autonomous driving or video surveillance. However, its performance may not be as good as more complex two-stage detectors on some datasets with small or heavily occluded objects.

On the other hand, a typical CNN model may be more suitable for image classification tasks where the goal is to predict the correct label for an entire image, rather than detecting and localizing specific objects within the image. Therefore, choose the yolo5 model for the car license task is more suitable than a typical CNN model.

In [ ]: